The initial development of transmission control protocol (TCP) was based on networking and processing capabilities that were then currently available. As a result, various fundamental assumptions regarding its operation were prefaced on networking and processor technologies that existed at that time. Among the assumptions on which TCP was prefaced includes the scarcity and high cost of bandwidth and the partially limitless processing resources available by a host processor. With the advent of technologies such as Gigabit Ethernet (GbE), these fundamental assumptions have radically changed to the point where bandwidth is no longer as scarce and expensive and the host processing resources are now regarded a being limited rather than virtually infinite. In this regard, the bottleneck has shifted from the network bandwidth to the host processing bandwidth. Since host processing systems do more than merely providing faster network connections, shifting network resources to provide much faster network connections will do little to address the fundamental change in assumptions. Notably, shifting network resources to provide much faster network connections would occur at the expense of executing system applications, thereby resulting in degradation of system performance.
Although new networking architectures and protocols could be created to address the fundamental shift in assumptions, the new architectures and protocols would still have to provide support for current and legacy systems. Accordingly, solutions are required to address the shift in assumptions and to alleviate any bottlenecks that may result with host processing systems. A transmission control protocol offload engine (TOE) may be utilized to redistribute TCP processing from the host system onto specialized processors which may have suitable software for handling TCP processing. The TCP offload engines may be configured to implement various TCP algorithms for handling faster network connections, thereby allowing host system processing resources to be allocated or reallocated to application processing.
In order to alleviate the consumption of host resources, a TCP connection can be offloaded from a host to a dedicated TCP/IP offload engine (TOE). Some of these host resources may include CPU cycles and subsystem memory bandwidth. During the offload process, TCP connection state information is offloaded from the host, for example from a host software stack, to the TOE. A TCP connection can be in any one of a plurality of states at a given time. To process the TCP connection, TCP software may be adapted to manage various TCP defined states. Being able to manage the various TCP defined states may require a high level of architectural complexity in the TOE.
Offloading state information utilized for processing a TCP connection to the TOE may not necessarily be the best solution because many of the states such as CLOSING, LAST_ACK and FIN_WAIT—2 may not be performance sensitive. Furthermore, many of these non-performance sensitive states may consume substantial processing resources to handle, for example, error conditions and potentially malicious attacks. These are but some of the factors that substantially increase the cost of building and designing the TOE. In addition, a TOE that has control, transferred from the host, of all the state variables of a TCP connection may be quite complex, can use considerable processing power and may require and consume a lot of TOE onboard-memory. Moreover, the TCP connection offloaded to the TOE that has control, transferred from the host, of all the state variables of the TCP connection can be inflexible and susceptible to connection loss.
TCP segmentation is a technology that may permit a very small portion of TCP processing to be offloaded to a network interface card (NIC). In this regard, a NIC that supports TCP segmentation does not truly incorporate a full transmission control processing offload engine. Rather, a NIC that supports TCP segmentation only has the capability to segment outbound TCP blocks into packets having a size equivalent to that which the physical medium supports. Each of the outbound TCP blocks is smaller than a permissible TCP window size. For example, an Ethernet network interface card that supports TCP Segmentation, may segment a 4 KB block of TCP data into 3 Ethernet packets. The maximum size of an Ethernet packet is 1518 bytes inclusive of header and a trailing CRC.
A device that supports TCP segmentation does track certain TCP state information such as the TCP sequence number that is related to the data that the offload NIC is segmenting. However, the device that supports TCP segmentation does not track any state information that is related to inbound traffic, or any state information that is required to support TCP acknowledgements or flow control. A NIC that supports full TCP offload in the established state is responsible for handling TCP flow control, and responsible for handling incoming TCP acknowledgements, and generating outbound TCP acknowledgements for incoming data.
TCP segmentation may be viewed as a subset of TCP offload. TCP segmentation allows the protocol stack or operating system to pass information in the form of blocks of TCP data that has not been segmented into individual TCP packets to a device driver. The block of data may be greater than the size of an Ethernet packet. For instance, the block of data to be segmented could be 4 Kbytes or 16 Kbytes. A network adapter associated with the device driver may acquire the blocks of TCP data, packetize the acquired blocks of TCP data into 1518-byte Ethernet packets and update certain fields in each incrementally created packet. For example, the network adapter may update a corresponding TCP sequence number for each of the TCP packets by incrementing the TCP sequence number for each of the packets. In another example, an IP identification (IP ID) field and flag field would also have to be updated for each packet. One limitation with TCP segmentation is that TCP segmentation may only be done on a block of data that is less than a TCP window size. This is due to the fact that a device implementing TCP segmentation has no influence over TCP flow control. Accordingly, the device implementing TCP flow control only segments outbound TCP packets.
A TCP segmentation device does not examine incoming packets and as such, has no influence over flow control. Any received acknowledgement packet is passed up to the host for processing. In this regard, acknowledgement packets that are utilized for flow control are not processed by the TCP segmentation device. Moreover, a TCP segmentation device does not perform congestion control or “slow-start” and does not calculate or modify any variables that are passed back to the operating system and/or host system processor.
Another limitation with TCP segmentation is that information tracked by TCP segmentation is only information that is pertinent for the lifetime of the TCP data. In this regard, for example, the TCP segmentation device may track TCP segmentation numbers but not TCP acknowledgement (ACK) numbers. Accordingly, the TCP segmentation device tracks only a minimal subset of information related to corresponding TCP data. This limits the capability and/or functionality of the TCP segmentation device. A further limitation with TCP segmentation is that a TCP segmentation device does not pass TCP processed information back to an operating system and/or host processor. This lack of feedback limits the TCP processing that otherwise may be achieved by an operating system and/or host system processor.
Further limitations and disadvantages of conventional and traditional approaches will become apparent to one of skill in the art, through comparison of such systems with some aspects of the present invention as set forth in the remainder of the present application with reference to the drawings.